Preview of VLBI/Gaia Data Release 1 differences

On 2016.09.14 Gaia DR1 catalogue of 1142679769 objects was released. The data are accessible here, 0.5 Tb after uncompressing.

A detailed invetigation of VLBI/"Gaia differences can be found in
https://arxiv.org/abs/1611.02630 and
https://arxiv.org/abs/1611.02632 and

Text below is now obsolete. If you really want to read, you can scroll it down.

I ran cross matching Gaia catalogue against the modern VLBI position catalogue. I found 6550 sources within 3" search radius and 6054 sources within 100 mas. Roughly speaking, 1/2 VLBI sources has a counterpart with Gaia. Some of these 502 sources with position differences in a range of [0.1, 3] arcsec may be wrong associations with stars, some of them may be caused by differences in the center of radio and optic emission. I restricted further analysis to 6054 sources with position differences in a range of [0, 100] mas. Of them, roughly one thousand, i.e. 1/6 have position differences in a range of [8, 100] mas and 5083 objects have position differences less than 8 mas.

Here is the plot of the distribution of 5083 radio sources that have a counterpart with Gaia within 8mas:

(Exclusion near the Galactic plane (purple line) is a feature of Gaia DR1).

Here is a plot of the normalized distribution of position differences:

Here green thick line represents normalized distribution of the arcs between Gaia and VLBI position and blue line shows the best fit to the Rayleighian distribution. As we see, Rayleighian distribution that corresponds to a case when position differences are caused by random errors along right ascension and declination that are perfectly Gaussian, does not fit well. The average position distance is 2.4 mas. Interesting, the average position distance is not changed much if to restrict analysis of the northern hemisphere (δ > -35°) where VLBI positions are mainly from VLBA (2.2 mas) and the southern hemisphere where VLBI positions are dominated by single-band LBA (2.6 mas). Here is the tail of the position distance distribution:

Distribution of uncertainties

Here is a plot of the normalized distribution of reported semi-major axes of the error ellipses in VLBI and Gaia catalogues among 5083 sources with position errors less than 8 mas:

The distributions are very similar, though Gaia has longer tail.

The mean of semi-major error ellipse is 1.2 mas for VLBI and 5.4 for Gaia. If to restrict our analysis to the zone δ > -35°, the mean errors are 1.0 mas for VLBI and 5.8 mas for Gaia. If to restrict our analysis to the zone δ < -35°, the mean errors are 2.1 mas for VLBI and 3.5 mas for Gaia. It is not clear why Gaia position errors are smaller in the southern hemisphere.

Conclusions

Gaia is very competitive against VLBI. A cursory analysis like that cannot answer the question which catalogue, VLBI or Gaia is more precise. There is an indication that among matches, Gaia has more sources with position errors greater than VLBI, but this needs to be carefully examined.

There is a significant number (> 1000) of sources with large distances Gaia/VLBI. I think understanding the origin of these differences will be a hot topic in next several years.

Back to Leonid Petrov's discussion page.

Last update: 2016.09.15_10:52:49
Web page was prepared by Leonid Petrov ()